Spectral data for "Improved GAF and Siamese network based fresh waxy corn kernel force prediction using hyperspectral data"

The hyperspectral image data of fresh waxy corn kernels used in this study were collected in early September 2024 from major fresh corn production areas in Siping, Jilin Province, China. Three representative commercial varieties—Jingkenuo 2000, Jinnuo262, and Meiyu 27—were selected, and multiple kernel samples were collected for each variety to ensure data diversity and representativeness. To simulate the mechanical stress experienced by kernels during harvesting, a texture analyzer (BosinTech, China) was used to perform compression tests on the top of individual kernels using a flat metal probe. The compression was carried out at a constant speed of 0.5 mm/s, with force levels ranging from 2 to 12 N in 0.5 N increments. For each force level, a randomly selected batch of 100 kernels per variety was tested under ambient temperature conditions (20 °C). After compression, kernels were grouped, labeled, and sealed in plastic bags to prevent moisture loss, and subsequently used for hyperspectral image acquisition. Hyperspectral imaging was performed using the Specim IQ system (Specim, Finland) under standardized lighting conditions. To enhance visualization and support downstream analysis, RGB images were also captured for each sample. The dataset is organized by variety, with each sample folder containing the hyperspectral image file (hyperspectral.dat), header file (metadata.hdr), and corresponding RGB image (rgb.png). Preliminary image processing and spectral curve extraction were conducted using MATLAB, providing high-quality data for subsequent kernel stress prediction modeling.